A Cross-Layer FL-Based Clustering Protocol to Support Multicast Routing in IoT-Enabled MANETs With CF-mMIMO

作     者：Amalia, Amalia Pramitarini, Yushintia Perdana, Ridho Hendra Yoga Shim, Kyusung An, Beongku 

作者机构：Hongik Univ Grad Sch Dept Software & Commun Engn Sejong 30016 South Korea Hankyong Natl Univ Sch Comp Engn & Appl Math Anseong 17579 Gyeonggi South Korea Hongik Univ Dept Software & Commun Engn Sejong 30016 South Korea 

出 版 物：《IEEE ACCESS》 (IEEE Access)

年 卷 期：2025年第13卷

页      面：3881-3899页

核心收录：

基　　金：National Research Foundation of Korea (NRF) Grant - Korea Government Ministry of Science and ICT (MSIT) [NRF-2022R1A2B5B01001190] 

主　　题：Protocols Cross layer design Ad hoc networks Routing Clustering algorithms Physical layer Ultra reliable low latency communication Cost function Stability criteria Solid modeling Cell-free massive MIMO clustering cross-layer federated learning IoT-enabled mobile ad hoc network 

摘      要：This paper proposes a novel cross-layer federated learning (FL)-based clustering (CFLC) protocol to support multicast routing in internet of things (IoT)-enabled mobile ad hoc networks (MANETs) with cell-free massive multiple-input multiple-output (CF-mMIMO). The proposed CFLC protocol leverages cross-layer and FL approaches to enhance network stability and connectivity by optimizing cluster head (CH) selection and cluster formation. The cross-layer design integrates physical layer information such as mobility (speed and direction), position, channel capacity, and remaining energy, with network layer information (connectivity) to maximize the cost function value for cluster formation. We design the FL model to improve the clustering performance and satisfy future mobile network requirements. Specifically, during the CH selection step, FL can decide which nodes should be elected as CHs or cluster members (CMs) by using classification. In the cluster formation step, FL addresses a regression problem by optimizing the cost function weights for parameters such as mobility similarity, link quality, remaining energy, and channel capacity to decide which CH each node should follow. The simulation results show that the proposed CFLC protocol outperforms the benchmark protocols in terms of connectivity, scalability, and control overhead. Additionally, the results indicate that the CFLC protocol performs particularly well when using the reference point group mobility (RPGM) model, highlighting its advantage over the random waypoint (RWP) mobility model in maintaining network stability and connectivity.